Automated Visual Inspection of Aircraft Exterior Using Deep Learning

Abstract

Aircraft visual inspections, or General Visual Inspections (GVIs), aim at finding damages or anomalies on the exterior and interior surfaces of the aircraft, which might compromise its operation, structure, or safety when flying. Visual inspection is part of the activities of aircraft Maintenance, Repair and Overhaul (MRO). Specialists perform quality inspections to identify problems and determine the type and importance that they will report. This process is time-consuming, subjective, and varies according to each individual. The time that an aircraft stays grounded without flight clearance means financial losses. The main goal of this work is to advance the state-of-the-art of defect detection on aircraft exterior with deep learning and computer vision. We investigate improvements to the accuracy of dent detection. Besides, we investigate new classes of identified defects, such as scratches. We also plan to demonstrate that it is possible to develop a complete system to automate the visual inspection of aircraft exterior using images of the aircraft acquired by drones. We will use deep neural networks for the detection and segmentation of defective regions. This system will aid in the elimination of subjectivity caused by human errors and shorten the time required to inspect an aircraft, bringing benefits to its safety, maintenance, and operation.